Download Universal enveloping algebra

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Cartesian tensor wikipedia , lookup

Dual space wikipedia , lookup

Bra–ket notation wikipedia , lookup

Polynomial ring wikipedia , lookup

Structure (mathematical logic) wikipedia , lookup

Basis (linear algebra) wikipedia , lookup

Commutative ring wikipedia , lookup

Invariant convex cone wikipedia , lookup

Modular representation theory wikipedia , lookup

Fundamental theorem of algebra wikipedia , lookup

Representation theory wikipedia , lookup

Boolean algebras canonically defined wikipedia , lookup

Laws of Form wikipedia , lookup

Heyting algebra wikipedia , lookup

Linear algebra wikipedia , lookup

Geometric algebra wikipedia , lookup

History of algebra wikipedia , lookup

Universal enveloping algebra wikipedia , lookup

Exterior algebra wikipedia , lookup

Vertex operator algebra wikipedia , lookup

Clifford algebra wikipedia , lookup

Transcript
MATH 223A NOTES 2011
LIE ALGEBRAS
65
17. Universal enveloping algebras
Recall that, for an associative algebra A with unity (1), a Lie algebra structure on A
is given by the Lie bracket [ab] = ab − ba. Let L(A) denote this Lie algebra. Then L
is a functor which converts associative algebras into Lie algebras. Every Lie algebra L
has a universal enveloping algebra U(L) which is an associative algebra with unity. The
functor U is “adjoint” to the functor L. The universal enveloping algebra is defined by
category theory. The Poincaré-Birkoff-Witt Theorem gives a concrete description of the
elements of the elements of U(L) and how they are multiplied. There is also a very close
relationship with the multiplication rule in the associated Lie group.
17.1. Functors. I won’t go through the general definition of categories and functors
since we will be working with specific functors not general functors. I will just use vector
spaces over a field F , Lie algebras and associative algebras (always with unity) as the
main examples.
Definition 17.1.1. A functor from the category of vector spaces to the category of
associative algebras both over F is defined to be a rule F which assigns to each F -vector
space V an associative algebra F(V ) over F and to each linear map f : V → W , an
F -algebra homomorphism f∗ : F(V ) → F(W ) so that two conditions are satisfied:
(1) (idV )∗ = idF (V )
(2) (f g)∗ = f∗ g∗ .
Recall that an F -algebra is an algebra which is also a vector space over F so that multiplication if F -bilinear. An F -algebra homomorphism is a ring homomorphism which is
also F -linear. We say that the homomorphism is unital if it takes 1 to 1.
In short: a functor takes objects to objects and morphisms to morphism and satisfies
the two conditions listed above.
Example 17.1.2. The mapping A �→ L(A) is a functor from associative algebras to Lie
algebras. For this functor, f∗ = f for all F -algebra homomorphisms f : A → B. The
reason that this works is elementary:
f [a, b] = f (ab − ba) = f (a)f (b) − f (b)f (a) = [f (a), f (b)]
We say that f∗ is f considered as a homomorphism of Lie algebras L(A) → L(B). The
two conditions are obviously satisfied and this defines a functor.
Example 17.1.3. The forgetful functor F takes an associative algebra (or Lie algebra)
A to the underlying vector space. F is defined on morphisms by f∗ = f . Since F -algebra
homomorphisms are F -linear by definition, this defines a functor.
Exercise 17.1.4. Show that the following diagram commutes.
AssAlg�
L
��
��
��
��
F
�
� LieAlg
�
��
��
�
�
��� F
V ect
66
MATH 223A NOTES 2011 LIE ALGEBRAS
17.2. Tensor and symmetric algebras. These are two important algebras associated
to any vector space. They are both graded algebras.
Definition 17.2.1. A graded algebra over F is an algebra A together with a direct sum
decomposition:
A = A0 ⊕ A1 ⊕ A2 ⊕ · · ·
so that Ai Aj ⊆ Ai+j . If A has unity (1) it should be in A0 . Elements in An are called
homogeneous of degree n.
Example 17.2.2. The polynomial ring P = F [X1 , · · · , Xn ] is a graded ring with P k
being generated by degree k monomials. The noncommutative polynomial ring Q =
F �X1 , · · · , Xn � is also a graded ring with Qk being generated by all words of length k
in the letters X1 , · · · , Xn . An example of a graded Lie algebra is the standard Borel
subalgebra B of any semisimple Lie algebra L. Then B 0 = H is the CSA and B k is the
direct sum of all Bβ where β has height k.
�
�
Exercise 17.2.3. Show that P k has dimension n+k−1
. For example, for n = 2, dim P k =
k
k + 1 with basis elements xi y k−i for i = 0, · · · , k. Qk has dimension nk .
Definition 17.2.4. Given a vector space V , the tensor algebra T (V ) is defined to be the
vector space
T (V ) = F ⊕ V ⊕ V ⊗ V ⊕ V ⊗ V ⊗ V ⊕ · · ·
with multiplication defined by tensor product (over F ). This is an associative graded
algebra with T k (V ) = V ⊗k , the k-fold tensor product of V with itself. Note that, in
degree 1, we have T 1 (V ) = V .
If V = F n then T (F n ) ∼
= F �X1 , · · · , Xn �. For example, V ⊗ V is n2 dimensional with
basis given by ei ⊗ ej . The tensor algebra has the following universal property.
Proposition 17.2.5. Any linear map ϕ from a vector space V to an associative algebra
A with unity extends uniquely to a unital algebra homomorphism ψ : T (V ) → A:
i
� T (V )
��
��
∃!ψ
�
∀ϕ �� �
�
V ��
A
Proof. ψ is must be given by ψ(1) = 1, ψ(v1 ⊗ v2 ⊗ · · · ⊗ vk ) = ϕ(v1 )ϕ(v2 ) · · · ϕ(vk ).
�
This proposition means that T (V ) is the universal associative algebra with unity generated by V .
Definition 17.2.6. The symmetric algebra S(V ) generated by V is defined to be the
quotient of T (V ) by the ideal generated by all elements of the form x ⊗ y − y ⊗ x. This
makes S(V ) into a commutative graded algebra with unity. Since the relations are in
degree 2, the degree 1 part is still the same: S 1 (V ) = T 1 (V ) = V .
For example, when V = F n we have S(F n ) ∼
= F [X1 , · · · , Xn ].
MATH 223A NOTES 2011
LIE ALGEBRAS
67
Proposition 17.2.7. Any linear map ϕ from a vector space V to a commutative algebra
A with unity extends uniquely to a unital algebra homomorphism ψ : S(V ) → A:
i
� S(V )
��
��
∃!ψ
�
∀ϕ �� �
�
V ��
A
17.3. Universal enveloping algebra. Following tradition, we define this by its desired
universal property.
Definition 17.3.1. For any (possibly infinite dimensional) Lie algebra L, the universal
enveloping algebra of L is defined to be any pair (U, i) where U is an associative algebra
with unity and i : L → L(U ) is a Lie algebra homomorphism with the property that, for
any other associative algebra with unity A and any Lie algebra homomorphism ϕ : L →
L(A) there is a unique unital algebra homomorphism ψ : U → A so that the following
diagram commutes where ψ∗ = ψ considered as a homomorphism of Lie algebras.
L ��
i
�
L(U )
��
��
ψ∗
�
ϕ ��
� �
L(A)
Example 17.3.2. One important example is the case when A = EndF (V ) is the algebra
of F -linear endomorphisms of a vector space V . Then L(A) = gl(V ) and ϕ : L → L(A) =
gl(V ) is a representation of L making V into an L-module. The algebra homomorphism
ψ : U → A = EndF (V ) makes V into a module over the associative algebra U . Therefore,
a module over L is the same as a module over U .
Proposition 17.3.3. The universal enveloping algebra (U, i) of L is unique up to isomorphism if it exists.
Proof. If there is another pair (U � , i� ) then, by the universal property, there are algebra
homomorphisms ψ : U → U � and ψ � : U � → U so that i� = ψ∗ i and i = ψ∗� i� . But then
i = ψ∗� ψ∗ i = (ψ � ψ)∗ i. By uniqueness, we must have ψ � ψ = idU . Similarly ψψ � = idU � . So,
U∼
�
= U � and i, i� correspond under this isomorphism.
The construction of U is easy when we consider the properties of an arbitrary Lie
algebra homomorphism
ϕ : L → L(A)
Since ϕ is a linear mapping from L to A, it extends uniquely to a unital algebra homomorphism ϕ : T (L) → A. Taking into account that ϕ is also a Lie algebra homomorphism,
we see that, for any two elements x, y ∈ L, we must have
ϕ([x, y]) = [ϕ(x), ϕ(y)] = f (x)f (y) − f (y)f (x) = ϕ(x ⊗ y − y ⊗ x)
In other words, ϕ : T (L) → A has the elements
x ⊗ y − y ⊗ x − [x, y]
68
MATH 223A NOTES 2011 LIE ALGEBRAS
in its kernel. Let J be the two-sided ideal in T (L) generated by all elements of this
form. Then J is in the kernel of ϕ and we have an induced unital algebra homomorphism
ψ : T (L)/J → A.
Definition 17.3.4. U(L) is defined to be the quotient of T (L) by the ideal generated by
all x ⊗ y − y ⊗ x − [x, y] for all x, y ∈ L. Let i : L → U(L) be the inclusion map i(x) = x.
Note that the relations imposed on U(L) are the minimal ones needed to insure that
i : L → L(U(L)) is a Lie algebra homomorphism. The fact that (U(L), i) satisfies the
definition of a universal enveloping algebra is supposed to be obvious.
Exercise 17.3.5. Show that, for any associative algebra A, there is a canonical unital
algebra homomorphism U(L(A)) → A. When is this an isomorphism? What happens
when A is commutative?
In the special case that L is a graded Lie algebra, such as a standard Borel algebra,
T (L) has another grading given by
�
T (L)k =
Lj1 ⊗ Lj2 ⊗ · · · ⊗ Ljm
P
ji =k
and the ideal J is generated by homogeneous elements. This makes U(L) into a graded
algebra. In general, there is no graded structure on U(L). However, there is a filtration
Uk (L) induced by the filtration
Tk (L) = T 0 (L) ⊕ T 1 (L) ⊕ · · · ⊕ T k (L)
of T (L).
By a filtration of an algebra A we mean a sequence of vector subspaces
A0 ⊆ A1 ⊆ A2 ⊆ · · ·
�
so that A = Ak and so that Aj Ak ⊆ Aj+k . The associated graded algebra is defined by
Gk (A) = Ak /Ak−1 with multiplication Gj Gk → Gj+k induced by the filtered multiplication
on A. We let G(L) denote the associated graded algebra of U(L).
Example 17.3.6. Suppose that L is abelian. Then U(L) is T (L) modulo the ideal
generated by x ⊗ y − y ⊗ x since [x, y] = 0. But this means U(L) is the symmetric algebra
S(L). If dim L = n then, in filtration k, we have Uk (L) = Sk (L) which is equivalent to the
vector space of polynomials of degree ≤ k in n variables, or equivalently, homogeneous
polynomials of degree equal to k in n + 1 variables. Thus
�
�
n+k
n
k
n+1
dim Uk (L) = dim Sk (F ) = dim S (F
)=
k
In this example, U(L) ∼
= G(L) ∼
= S(L).
MATH 223A NOTES 2011
LIE ALGEBRAS
69
17.4. Poincaré-Birkov-Witt. The PBW Theorem gives a detailed description of the
structure of the universal enveloping algebra U(L). It gives a formula for a vector space
basis and how they are multiplied. One of the amazing features of the theorem is that it
says that �the dimension
of Uk (L) depends only on the dimension of L. I.e., the dimension
�
n+k
is always k . For each k ≥ 0 consider the composition:
ϕk : T k (L) = L⊗k �→ Tk (L) � Uk (L) � G k (L) = Uk (L)/Uk−1 (L)
� k
Lemma 17.4.1. ϕ =
ϕ : T (L) → G(L) is a graded algebra epimorphism which
induced a graded algebra epimorphism S(L) � G(L).
Proof. The morphism ϕ is multiplicative by definition. So, it is an algebra epimorphism.
Elements of the form x ⊗ y − y ⊗ x ∈ T 2 (L) are sent to [x, y] ∈ U1 (L) which is zero in
G2 (L). Therefore, there is an induced algebra epimorphism S(L) � G(L).
�
Theorem 17.4.2 (PBW). The natural graded epimorphism S(L) → G(L) is always an
isomorphism.
Corollary 17.4.3 (PBW basis). If x1 , · · · , xn is a vector space basis for L then a vector
space basis for Uk (L) is given by all monomials of length ≤ k of the form
xj1 xj2 · · · xj�
where j1 ≤ j2 ≤ j3 ≤ · · · ≤ j� plus 1 (given by the empty word). In particular, i : L →
U(L) is a monomorphism.
Proof. Monomials as above (with nondecreasing indices) of length equal to k form a basis
for S k (L) and therefore give a basis for Uk (L) modulo Uk−1 .
�
Corollary 17.4.4. If H is any subalgebra of L then the inclusion H �→ L extends to a
monomorphism U(H) �→ U(L). Furthermore U(L) is a free U(H)-modules.
Proof. Extend an ordered basis of H to an ordered basis for L and use PBW bases.
�
We will skip the proof of the PBW Theorem in the lectures. So, you need to read
the proof. There is a purely algebraic proof in the book. Here I will give a proof using
combinatorial group theory.